Novel study designs and data technologies for environmental epidemiology

Abstract

Introduction. Availability of modern data technologies, such as wearables, remote sensors in satellites, and re-analysis modelling provide exceptional opportunities for environmental research. These tools allow the longitudinal collection of individual-level information, and the linkage with finely reconstructed spatio-temporal exposure maps. However, traditional analytical methods are not well suited in this complex data setting. In this contribution we illustrate the application of a new study design called case time series to analyse short-term association between environmental exposures and allergic symptoms in a smartphone study.Methods. Data were collected within AirRater, an integrated online platform operating in Tasmania. Daily events of allergic symptoms were collected from 1,601 subjects in 2015-2018 through a smartphone app. Geolocation allowed the linkage with spatio-temporal measures of pollen (grains/m3), fine particulate matter (PM2.5, μg/m3), and temperature (Celsius) from ground stations and re-analysis models. Individual outcome and exposure series were analysed with a case time series design, fitting conditional Poisson models with distributed lag models to estimate dependencies while enforcing a strict temporal control through subject/month strata intercepts, natural splines of time, and indicators of day of the week.Results. We found increased risk of allergic symptoms associated independently with all the three environmental factors. Pollen shows a step increase in risk that flattens out at high exposures, and a lagged effect up to two days. Risks of PM2.5, is linear and mostly limited to the same-day exposure. Temperature displays non-linear associations, with increases in allergic symptoms beyond daily averages of 15°C.Discussion. The combination of novel study designs and modern data technologies allows investigation of complex epidemiological relationships using individual-level longitudinal data and ensuring strict control for time-invariant and time-varying factors. This flexible modelling framework can be adapted to various contexts and research areas.